In order to improve the learning performance of neural networks with small amounts of samples so as to enhance the learning speed

a novel artificial affective neural network (AANN) is explicitly introduced in this paper. Traditional neural networks are integrated with non-fully connected affective neurons whose non-linear activation functions are different from traditional ones. Using two affective factors termed as “nervousness” and “confidence” to simulate the emotional changes of human brains

the cognitive ability of neural networks can be practically increased

which contributes to the speed of learning. A numerical example and the superheated steam temperature predictive control strategy of boilers concerning AANN applications are demonstrated

showing better performances compared with those of BP neural networks and support vector machine (SVM).